WAON User Interview: Yamaha Corporation "Without WAON, the software enabling large-scale model analysis, and without our technology for feeding back the WAON analysis results into product development, this development would not have been possible." -WAON large-scale acoustic analysis was instrumental in development of an audio system based on the world's first^(*1) HRTF with full 24kHz coverage-

Introduction

From left to right:
Y. Shiozawa, A. Miki, M. Katayama,
H. Onitsuka, T. Suenaga

The interviewees
Yamaha Corporation Center for Advanced Sound Technologies Innovative Technology Division Senior Engineer Hirofumi Onitsuka Assistant Manager Akira Miki Tsukasa Suenaga Yasuo Shiozawa Technology Development Group2 Technology Development Department AV Products Division Assistant Manager Masaki Katayama

For this report, we called on Yamaha Corporation. As a manufacturer of all kinds of musical instruments, the company continues to lead the market. Its high-end AV products, incorporating audio technology accumulated over the years, are also big hits with audio fans throughout the world.

Now comes the DVX-1000, released in March 2008 as the world's first 24kHz HRTF based home theater system. The development of this system represents the culmination of Yamaha's audio technology and acoustic simulation know-how. In this series of interviews, we talked with Yamaha people about the back-story of the development project, and about the use of the large-scale acoustic analysis software WAON in the project.

Interview Contents

First tell us about the work you do at Yamaha.

Figure 1. DVX-1000 *Received 5 star review from What Hi-Fi Sound and Vision (Haymarket Consumer Media Ltd) ( DVD Home Theater System DVX-1000 Website )

What can you tell us about the next-generation virtual surround technology adopted in the DVX-1000?

Katayama:

Figure 2. Schematic of next-generation virtual sound technology AIR SURROUND XTREME Masaki Katayama When you go to a movie theater, there are speakers all around you. The sound from these speakers gives the moviegoer the sensation of being immersed right in the world on the screen. This is what is called surround. A home theater system applies the same principle, surrounding the viewer with speakers. A 5.1ch system, the most popular kind today, uses one front-center speaker, two front speakers (left and right), two surround speakers to the left and right of the viewer, and also a sub-woofer, for a total of six speakers. A more advanced 7.1ch system adds two speakers in the rear. Such a setup will not always be practical in the home, however. There may not be enough space, or maybe you don't want all those speaker cables in the way when doing housecleaning. So this development project started out by asking if we could somehow achieve a similar effect using virtual technology. With the DVX-1000, we succeeded in realizing sound equivalent to that of a 7.1ch high-end home theater system, using only two center-front speakers and a subwoofer [Figure 2]. That's something no one else in the world has accomplished besides Yamaha.

Could you describe the course of the development and features of the product?

Katayama:

Spatial sound pressure distribution near head Spatial sound pressure distribution near ears Sound pressure distribution on head surface Figure 3. Head-related transfer function (HRTF) calculation Methodology: WAON fast multi-pole boundary element method (FMBEM) The project started 10 years ago. At that time, virtual surround itself was already available in a number of products, so we decided to focus on 7.1ch, which would set us apart from the competition in a huge way. With the DVX-1000, sound really seems to be coming from the rear. Even surround systems that actually have speakers in the rear can't always achieve this effect, so ordinarily it would be unthinkable in a virtual system. The key to creating such realistic sound is the accurate quantification of the HRTF (head-related transfer function), which represents how sound emanating from various directions arrives at the left and right eardrums. People point to two major issues with conventional virtual surround technology. The first is that the HRTF used for the analysis is based on a onesize- fits-all model of the head, with insufficient allowance made for individual differences in head shapes of listeners. So we ran simulations for lots of different head models [Figure 3]. On that basis we developed a new, highly universal head model, which in turn enabled us to obtain a next- generation HRTF, which means that all kinds of listeners can experience realistic sound localization - the direction from which sound arrives, and true depth perception. The important point here is that we used simulations, not actual measurements. For actual measurements, dummy heads are used for sound recording; but it wouldn't be practical to create dummy heads accounting for the many individual variations in head shape. Simulation, on the other hand, let us create a large number of head shapes, which could be analyzed immediately for calculating HRTF. That's why we were able to develop a head model that allowed us to calculate an HRTF applicable to just about everyone. We asked people from various countries to listen to virtual surround using this HRTF. They were all astounded to hear sound coming from behind them. Seeing their reaction convinced us that this model is one we could use. It took us three years from the time we began the development using simulations, but at last we succeeded. A second issue is that the HRTF used up to now could not cover the full 20kHz spectrum audible to the human ear. The actual range of frequencies that people can hear as "sound" is said to be a bit over 15kHz, so you might wonder why it's necessary to go to that extreme. Realizing virtual surround, however, requires performing calculations with sound data, and there is a big difference in the results when data used for the calculations is all guaranteed over the necessary range and when the full range is not guaranteed. When calculations are made using non-guaranteed data, some kind of human intervention takes place in the nonguaranteed range. We believe that this intervention is what causes the sound quality deterioration typical of virtual surround, such as live instruments not sounding convincingly live, or the human voice having a mechanical sound. Thanks to WAON, however, we were able to obtain precise data guaranteed over the full range including high frequencies, so no intervention was necessary. As a result, we were able to reproduce sound fields with very high accuracy. The DVX-1000 features what is called a music mode, which truly makes you feel you are surrounded by live instruments. This is technology only Yamaha has, and it's winning high praise from customers. These achievements were the culmination of our many years of experience with acoustic technology and simulation know-how, but this alone would not have been enough. Without WAON, the software enabling large-scale model analysis, and without our technology for feeding back the WAON analysis results into product development, this development would not have been possible. Both these technologies together were what made this development happen.

Onitsuka:

Hirofumi Onitsuka The concept has been around for quite some time, and we could just about predict the outcome from our experience. But there was no tool for actually carrying it out. We tried other tools, but the best we could do is analyze up to around 12 kHz . Going beyond that seemed out of the question. Then Cybernet introduced us to WAON, adopting FMBEM (fast multi-pole boundary element method), and we thought, this is just what we were looking for. The DVX-1000 we've been talking about is what came into being as a result.

What plans do you have for the future?

Our structural-acoustic coupled analysis function was released in September 2008, so we hope you will make use of it. Do you have any other comments on WAON functions?

Structural - Acoustic coupled analysis function

Suenaga:	Tsukasa Suenaga Speed of analysis is one thing, but keeping memory use within practical limits is the big issue. Before calculation time becomes an issue, we can't even begin if there isn't enough memory.
Onitsuka:	The kinds of sounds Yamaha deals with are mostly music, so we have to cover the whole range from lows to highs. That tends to result in large amounts of data to be processed.
Suenaga:	What struck me when I tried it out was how simple it is. There are products from outside of Japan with lots of functions, but sometimes with all those choices you can wind up in a dead-end. There's no such worry with WAON.
Onitsuka:	When we train CAE engineers in our company, the most important thing is getting off to a proper start. When a user looks at a software GUI or manuals and doesn't feel entirely comfortable with it right away, that generally means it's going to take longer to learn the software, and the desire to learn it will be diminished. With WAON, however, nearly everyone gets into it immediately. You feel like, Hey, this is something even I can do. This is really helpful from a teaching standpoint. Up to now, we've had to teach them how to do everything, from the basic details; whereas with WAON they can pretty much get started all by themselves, without having to be told how.
Miki:	Akira Miki I was pretty much a beginner when it comes to acoustic analysis, but it was easy to get up and running right from t h e beginning. There weren't a lot of choices to worry about; you just proceed all the way to the end and the results come out. Not only that, but the analysis results were right in line with empirical results up to now, so I don't think reliability is a problem.

Our development goal from the start was to design the software for intuitive use, without having to read a manual. Even if we add more functionality to it, we'd like to make sure we don't sacrifi ce ease of use. Do you have any other requests you'd like to make?

Onitsuka:	I would like to see a lot more plus-alpha features that no other competing products have, such as the ability to include in analyses the components that are left out in the case of the conventional wave equation model. It would also be useful if there were a simple way to compare experimental data and analytical data. Enhanced customization functions would be helpful as well.
Shiozawa:	Yasuo Shiozawa I would like to see a bit more functionality for representing results, such as animation or cross-sectional views and the like.
Onitsuka:	A more intuitively understandable way of displaying vector coeffi cient results would be nice, too.

Thank you for sharing your views with us.
We intend to continue reflecting customer needs in our on-going development of WAON features. What about Cybernet? Is there anything you would like us to do better?

We still have a long way to go, but I believe you are exactly right. We hope for your continued support as we try to put such ideals into practice. Our sincere appreciation to Hirofumi Onitsuka, Masaki Katayama, Akira Miki, Tsukasa Suenaga and Yasuo Shiozawa of Yamaha Corporation for taking time from their busy schedules to sit for this interview.